Social Bot Detection using Variational Generative Adversarial Networks with Hidden Markov Models in Twitter Network
Abstract
In online social networks (OSNs), bots are creative software programs which attempt to manipulate and influence benign accounts by disseminating fake information. Recent approaches to bot identification suffer from imbalanced data, as well as overfitting and scalability related issues. To address these challenges, this paper proposes a novel two-stage approach, called hidden Markov model based variational generative adversarial network (HMM-VGAN). The first stage trains the variational autoencoder (VAE) using a set of features, such as profile, URL, word embedding over linguistic, user behavioral pattern, and social relationship with a goal to remove noisy data. The second stage integrates VAE with generative adversarial networks (VGAN) which is used for generating an augmented data to overcome imbalanced distribution of data. The second stage proposes an ensemble based HMM with a pre-trained recurrent neural network of VGAN algorithm for bot identification. Experimental results using three Twitter datasets demonstrate the efficacy of the proposed HMM-VGAN algorithm in terms if performance metrics such as precision, recall and accuracy. Precisely, the proposed HMM-VGAN algorithm achieves around 95% precision and up to 4%–10% improvement on precision over other existing approaches.
Recommended Citation
G. Lingam and S. K. Das, "Social Bot Detection using Variational Generative Adversarial Networks with Hidden Markov Models in Twitter Network," Knowledge-Based Systems, vol. 311, article no. 113019, Elsevier, Feb 2025.
The definitive version is available at https://doi.org/10.1016/j.knosys.2025.113019
Department(s)
Computer Science
Keywords and Phrases
Bot; Generative adversarial networks; Hidden Markov model; Variational auto-encoder
International Standard Serial Number (ISSN)
0950-7051
Document Type
Article - Journal
Document Version
Citation
File Type
text
Language(s)
English
Rights
© 2025 Elsevier, All rights reserved.
Publication Date
28 Feb 2025
